IT (Intermediary Guidelines) Amendment Rules, 2026 — Synthetic Media
Applies to: Intermediaries and Significant Social Media Intermediaries (SSMIs); platforms enabling creation or dissemination of synthetic/AI-generated content
India's first statutory framework for synthetically generated information (SGI) — AI-created or AI-altered audio-visual content, including deepfakes. Notified by MeitY on 10 February 2026 under Section 87 of the IT Act and effective from 20 February 2026, it amends the IT (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021. SGI is defined broadly as information artificially created, generated, modified or altered using a computer resource in a manner that appears authentic. Core obligations: intermediaries enabling SGI creation must label such content prominently (a visible label covering at least 10% of visual surface area, or the first 10% of audio duration) and embed permanent, non-removable metadata identifying the originating computer resource. Significant Social Media Intermediaries must obtain user declarations on whether uploaded content is synthetic, deploy technical measures to verify those declarations, and label confirmed SGI. Takedown timelines are sharply compressed — unlawful content must be removed within 3 hours of a court or government order (down from 36 hours), with grievance resolution cut to 36 hours; particularly harmful categories (non-consensual intimate imagery, impersonation) carry windows as short as 2 hours. Failure to act can forfeit safe-harbour protection under Section 79. A follow-on draft amendment (21 April 2026) proposes continuous, always-visible labelling for the full duration of synthetic content.
Why it matters
This rule reaches further than its deepfake headline suggests. The forfeiture of safe harbour is the teeth — non-compliance is no longer a fine, it is the loss of the legal shield that makes a platform viable, which converts labelling and takedown from best-practice to existential. For any CII operator running a customer-facing platform with user-generated or AI-assisted content, the 3-hour clock (2 hours for the worst categories) is operationally brutal and effectively forces automated detection with human oversight layered on top. Watch the constitutional challenge risk: the compelled-labelling and pre-publication-verification provisions are drawing Article 19(1)(a) free-speech objections, so the operational detail may shift under litigation. The forward signal is provenance — permanent embedded metadata is now a legal expectation in India, and aligning your content pipeline to emerging provenance standards (C2PA-style) now is the no-regrets move.